LNA input protection that passes CW NF and fails on a blocker: the diode that spent your IIP3

Started by LunaWei · 2026-11-15 01:51 UTC · 23 views · 2 replies
L Admin
[Technical article / FAQ, rfforum Editorial Team]

An LNA that meets NF on a clean CW tone and collapses under a realistic blocker is often not a matching problem. Input protection diodes and clamp networks that look invisible at low power spend linearity and sometimes a slice of NF once the blocker turns them on.

1. What the clamp actually does
- A shunt diode or TVS across the input looks open at small signal and starts conducting as the blocker peaks. The effective impedance and intermod products then belong to the clamp, not the LNA core.
- CW NF with a noise source rarely exercises the clamp. A two-tone or blocker+noise test will.
- Bias networks that starve the clamp recovery after a burst invent a "hang" that looks like AGC or front-end desense.

2. Freeze list before you retune the matching network
- Blocker level, frequency offset, and whether NF/IIP3 were measured with the clamp populated and with it lifted
- Scope or spectrum evidence that the clamp conducts at the failing level
- Whether a higher-threshold clamp or a series limiter topology restores IIP3 without breaking ESD
- Cable and generator: confirm the blocker at the LNA plane, not only at the source

3. Habit
Never close a "NF OK, blocker fail" ticket without a clamp on/off or threshold A/B. Archive blocker level, IIP3, and clamp PN next to every desense claim. If lifting the clamp kills the IMD, the matching network was never the patient.

4. Practical gate
- Measure NF and a simple two-tone with clamp in and out
- Write the blocker level that first moves IIP3 by 1 dB
- Prefer a clamp that stays off at your max expected interferer with margin

Discussion: on your last blocker desense, did the fix land in clamp threshold, bias recovery, or a real LNA stage — and do you forbid shipping NF-only acceptance for protected front ends?

— rfforum 编辑部 / Editorial Team
L Admin
Addendum — distinguish this from a limiter/ESD “NF tax” ticket. The question here is not only how many tenths of a dB the clamp costs on a clean CW NF, but at what blocker level the clamp starts conducting and spends IIP3. A front end that ships on NF-only acceptance will pass in the lab and then fail in the field.

A cheap gate: NF and a simple two-tone or blocker-plus-noise, clamp populated and clamp lifted (or a higher-threshold part), with the blocker measured at the LNA plane — not only at the generator. Write the blocker level that first moves IIP3 by 1 dB. If lifting the clamp kills the IMD, retuning the match was never the fix. Check bias recovery after a burst so a clamp “hang” is not logged as AGC desense.

Open: on your last blocker desense, did the fix land in clamp threshold, bias recovery, or a real LNA stage — and do you forbid shipping NF-only acceptance for protected front ends?

— rfforum 编辑部 / Editorial Team
N Member
Ran into something similar last year. Turned out to be a power integrity issue. Worth checking first.
广告 AdRenesas 瑞萨微控制器芯片产品照片Renesas Electronics微控制器、模拟与射频/时钟方案 · MCUs, analog, RF and timing访问官网 →免费展示 · 资料来自公开信息 · 如需撤下请联系 ads@rfforum.com
广告 Ad广告位招商 / Advertise hereads@rfforum.com · 刊例 Rate cardT2 · 728×90
Log in or register to reply.